JP2009059048A - Image processor, imaging device, image processing method, and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To perform more reliable image detection at high speed. <P>SOLUTION: When an image to be detected is detected on an input image by means of identification action using a reference image, a storage unit 250 stores a normal reference image as well as a contracted reference image such that a feature portion shown by the normal reference image is contracted. An image conversion unit 211 contracts the input image stored in an image memory 230. A second discriminator 215 performs a discriminating action using the contracted reference image over the image contracted by the image conversion unit 211. A discrimination result determining unit 216 specifies the position of the target image that has been detected on the contracted image to be detected. A discrimination area setting unit 213 detects the image to be detected on the input image by performing a discriminating action using the normal reference image, at a position on the input image that corresponds to the position specified. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to an image processing device, an imaging device, an image processing method, and a program, and more particularly, to an image processing device, an imaging device, an image processing method, and a program suitable for detecting a specific image portion.

  In an imaging apparatus such as a digital still camera, a technique for detecting a face on a viewfinder screen when a person is a subject is realized (for example, Patent Document 1).

  By using such face detection, for example, it is possible to focus on the face part, and to adjust exposure and white balance according to the face part, making it easy to shoot optimized for a person. Can do.

In such a face detection technique, a plurality of images (so-called image pyramids) obtained by sequentially reducing an input image are generated, and a discrimination operation using a face image (reference image) prepared in advance is performed on each image. Even if the size of the face portion in the captured image is indefinite, if a reference image having a certain size is prepared, face detection can be performed.
JP 2003-36439 A

  In the face detection by the conventional image pyramid, an image pyramid obtained by sequentially reducing the input image is used. Therefore, it is possible to detect a face larger than the size of the face indicated by the reference image. If it appears small, it cannot be detected.

  For this reason, for example, when faces of various sizes appear on the imaging apparatus by imaging a plurality of subjects having different distances from the imaging position, a small face may not be detected.

  In order to eliminate such inconvenience, it is conceivable to perform face detection without reducing the input image. However, in the method of performing face identification while moving the sub-window on the image, the processing man-hours increase as the image size increases. Will increase. As a result, the processing time required for detection becomes long, and there is a possibility that the face cannot be detected in real time on the finder screen. As the performance of the imaging device improves, the input image tends to become larger and the display device has a higher definition, making it more difficult to perform reliable image detection in real time.

  SUMMARY An advantage of some aspects of the invention is that it provides an image processing apparatus, an imaging apparatus, an image processing method, and a program capable of performing more reliable image detection at high speed.

In order to achieve the above object, an image processing apparatus according to the first aspect of the present invention provides:
In an image processing apparatus for detecting a detection target image on an input image by an identification operation using a reference image,
Reference image storage means for storing the reference image;
Reduced reference image storage means for storing a reduced reference image obtained by reducing the characteristic portion indicated by the reference image stored in the reference image storage means;
Image reduction means for reducing the input image;
Target position specifying means for specifying the position of the detection target image in the reduced image by performing an identification operation using the reduced reference image on the image reduced by the image reducing means;
Image detecting means for detecting the detection target image on the input image by performing an identification operation using the reference image at a position on the input image corresponding to the position specified by the target position specifying means. When,
It is characterized by providing.

In the image processing apparatus,
It is desirable that the image detection unit does not perform an identification operation on the input image when the target position specifying unit cannot specify the position of the detection target image on the reduced image.

In the image processing apparatus,
The image reduction means sequentially reduces the input image,
It is desirable that the image detection unit detects the detection target image by performing an identification operation using the reference image in each of the images sequentially reduced by the image reduction unit.

In order to achieve the above object, an imaging apparatus according to a second aspect of the present invention provides:
In an imaging device that detects a detection target image on an input image by an identification operation using a reference image,
Input image acquisition means for acquiring a captured image as an input image;
Reference image storage means for storing the reference image;
Reduced reference image storage means for storing a reduced reference image obtained by reducing the characteristic portion indicated by the reference image stored in the reference image storage means;
Image reduction means for reducing the input image;
Target position specifying means for specifying the position of the detection target image in the reduced image by performing an identification operation using the reduced reference image on the image reduced by the image reducing means;
Image detecting means for detecting the detection target image on the input image by performing an identification operation using the reference image at a position on the input image corresponding to the position specified by the target position specifying means. When,
It is characterized by providing.

In the imaging apparatus,
It is desirable that the image detection unit does not perform an identification operation on the input image when the target position specifying unit cannot specify the position of the detection target image on the reduced image.

In the imaging apparatus,
The image reduction means sequentially reduces the input image,
It is desirable that the image detection unit detects the detection target image by performing an identification operation using the reference image in each of the images sequentially reduced by the image reduction unit.

In order to achieve the above object, an image processing method according to a third aspect of the present invention includes:
An image processing method for detecting a detection target image on an input image by an identification operation using a reference image,
An image reduction step for reducing the input image;
An object for specifying the position of the detection target image in the reduced image by performing an identification operation using the reduced reference image obtained by reducing the characteristic portion indicated by the reference image on the image reduced in the image reduction step. A location step;
An image detection step of detecting the detection target image on the input image by performing an identification operation using the reference image at a position on the input image corresponding to the position specified in the target position specifying step. When,
It is characterized by including.

In order to achieve the above object, a program according to the fourth aspect of the present invention is:
A program for causing a computer to detect a detection target image on an input image by an identification operation using a reference image,
In the computer,
A function of reducing the input image;
A function of identifying the position of the detection target image in the reduced image by performing an identification operation using the reduced reference image obtained by reducing the feature portion indicated by the reference image on the reduced image;
A function for detecting the detection target image on the input image by performing an identification operation using the reference image at a position on the input image corresponding to the specified position;
It is characterized by realizing.

  According to the present invention, more reliable image detection can be performed at high speed.

  Embodiments according to the present invention will be described below with reference to the drawings. In the present embodiment, a case where the present invention is realized by a digital still camera (hereinafter referred to as a digital camera) is illustrated. The digital camera 1 according to the present embodiment has a function of detecting a specific subject portion on the finder screen and performing an imaging setting in accordance with the detected subject portion, in addition to the function of a general digital still camera. It shall be. In this embodiment, it is assumed that the face of a person on the finder screen is detected as a subject part. That is, the digital camera 1 according to this embodiment is a digital still camera having a so-called face detection function.

  FIG. 1 is a block diagram showing a configuration of a digital camera 1 according to an embodiment of the present invention. The schematic configuration of the digital camera 1 according to the present embodiment includes an imaging unit 100, a data processing unit 200, an interface (I / F) unit 300, and the like as illustrated.

  The imaging unit 100 is a part that performs an imaging operation of the digital camera 1, and includes an optical device 110, an image sensor unit 120, and the like as illustrated.

  The optical device 110 includes, for example, a lens, a diaphragm mechanism, a shutter mechanism, and the like, and performs an optical operation related to imaging. In other words, the operation of the optical device 110 collects incident light and adjusts optical elements related to the angle of view, focus, exposure, and the like, such as focal length, aperture, and shutter speed. Note that the shutter mechanism included in the optical device 110 is a so-called mechanical shutter device, and when the shutter operation is performed by the image sensor, the optical device 110 may not include the shutter mechanism. The optical device 110 operates under the control of the control unit 210 described later.

  The image sensor unit 120 generates an electrical signal corresponding to the incident light collected by the optical device 110, for example, a CCD (Charge Coupled Device) or a CMOS (Complementary Metal Oxide Semiconductor). (Semiconductor) image sensor. The image sensor unit 120 performs photoelectric conversion to generate an electrical signal corresponding to the received light and output it to the data processing unit 200.

  When a liquid crystal screen (such as a display unit 310 to be described later) or EVF (Electronic View Finder) is used as a viewfinder of the digital camera 1, the image sensor unit 120 constantly outputs a low-resolution image signal. It is assumed that a so-called live view image (Rexro image) is displayed as a viewfinder screen. In the present embodiment, a display unit 310 to be described later is used as a finder screen, and a live view function is realized by displaying a Rexro image on the finder screen of the display unit 310 during an imaging operation with the digital camera 1. To do.

  Rexroux images displayed with such a live view function usually use display images that are relatively small in size (low-resolution images) so that the processing load is low because display speed is more important than image quality. In this embodiment, it is assumed that the image sensor unit 120 outputs a 640 × 480 VGA (Video Graphics Array) image except when imaging by a shutter operation, and this VGA image is used as a Rexro image for live view. To do. The size (resolution) of the Rexro image is defined according to the performance (resolution, etc.) of the display device that displays the live view image. In other words, the output display device has a resolution (size) enough to obtain image quality and display speed that can be used as a finder.

  The data processing unit 200 processes the electrical signal generated by the imaging operation by the imaging unit 100 to generate digital data indicating the captured image, and performs image processing on the captured image. As shown in FIG. 1, the data processing unit 200 includes a control unit 210, an image processing unit 220, an image memory 230, an image output unit 240, a storage unit 250, an external storage unit 260, and the like.

  The control unit 210 includes, for example, a processor such as a CPU (Central Processing Unit), a main storage device such as a RAM (Random Access Memory), and the like, and is stored in a storage unit 250 described later. By executing the above, each part of the digital camera 1 is controlled. Further, in the present embodiment, by executing a predetermined program, a function related to each process described later is realized by the control unit 210.

  The image processing unit 220 includes, for example, an ADC (Analog-Digital Converter), a buffer memory, an image processing processor (a so-called image processing engine), and the like. Based on the signal, digital data indicating a captured image is generated.

  That is, when the analog electric signal output from the image sensor unit 120 is converted into a digital signal by the ADC and sequentially stored in the buffer memory, the image processing engine performs so-called development processing on the buffered digital data, Adjust image quality and compress data.

  The image memory 230 includes, for example, a storage device such as a RAM or a flash memory, and temporarily stores captured image data generated by the image processing unit 220, image data processed by the control unit 210, and the like. In the present embodiment, it is assumed that image data output from the image sensor unit 120 is temporarily stored as a live view image for a finder, and processing for face detection is performed using the stored image data.

  The image output unit 240 includes, for example, an RGB signal generation circuit and the like, converts the image data expanded in the image memory 230 into an RGB signal and the like, and outputs the RGB signal to a display screen (a display unit 310 to be described later).

  The storage unit 250 includes a storage device such as a ROM (Read Only Memory) or a flash memory, and stores programs and data necessary for the operation of the digital camera 1. In the present embodiment, it is assumed that an operation program executed by the control unit 210 or the like is stored in the storage unit 250.

  The external storage unit 260 includes a storage device that can be attached to and detached from the digital camera 1 such as a memory card, and stores image data captured by the digital camera 1.

  The interface unit 300 has a configuration related to an interface between the digital camera 1 and its user or an external device. As shown in FIG. 1, the display unit 310, an external interface (I / F) unit 320, an operation unit 330, etc. Consists of

  The display unit 310 includes, for example, a liquid crystal display device, and displays and outputs various screens necessary for operating the digital camera 1, live view images at the time of shooting, captured images, and the like. In the present embodiment, display output of a captured image or the like is performed based on an image signal (RGB signal) from the image output unit 240.

  The external interface unit 320 includes, for example, a USB (Universal Serial Bus) connector, a video output terminal, and the like, and performs output of image data to an external computer device, display output of a captured image to an external monitor device, and the like. .

  The operation unit 330 includes various buttons configured on the outer surface of the digital camera 1, generates an input signal corresponding to an operation by the user of the digital camera 1, and inputs the input signal to the control unit 210. As buttons constituting the operation unit 330, for example, a shutter button for instructing a shutter operation, a mode button for designating an operation mode of the digital camera 1, a cross key and a function button for performing various settings, and the like Is included.

  Here, details of the image memory 230 of the data processing unit 200 will be described with reference to FIG. Here, the configuration of the image memory 230 necessary for processing related to the face detection function of the digital camera 1 will be described. That is, when the face detection function is executed by the digital camera 1, a storage area as shown in FIG. 2A is secured in the image memory 230 according to the present embodiment. As shown in the drawing, storage areas such as an input image storage area 230a, a detection image storage area 230b, and a detection image storage area 230c are prepared in the image memory 230, and image data corresponding to each storage area is stored. The

  The input image storage area 230a is a storage area for sequentially storing VGA images for live view output from the image sensor unit 120.

  The detection image storage area 230b is a storage area for storing a detection image for performing face detection related to the face detection function. In the present embodiment, a VGA image stored as a live view Rexro image (through image) in the input image storage area 230a is converted into a luminance image (hereinafter referred to as “detection through image PL”). It is stored in the image storage area 230b. An example of the detection through image PL stored in the detection image storage area 230b is shown in FIG.

  The detection image storage area 230c is a storage area for storing a detection image having a smaller size for performing face detection related to the face detection function. In the present embodiment, a detection through image PL stored in the detection image storage area 230b is reduced and converted to a smaller size (that is, a size smaller than VGA) (hereinafter referred to as a detection reduced image PS). Stored in the image storage area 230c. An example of the reduced detection image PS stored in the detection image storage area 230c is shown in FIG.

In the present embodiment, an image pyramid as used in the conventional face detection technology is created from the reduced detection image PS stored in the detection image storage area 230c. Therefore, for example, a 320 × 240 QVGA (Quarter-VGA) image is stored in the detection image storage area 230c as a size smaller than the VGA, and an image pyramid is created as shown in FIG. As the progress progresses, images obtained by further reducing the QVGA image are sequentially stored. Here, the detection reduced image PS of QVGA size is defined as a detection reduced image PS _0, and hereinafter, the detection reduced image PS ₁ , the detection reduced image PS ₂ ..., The detection reduced image according to the number of times of reduction. It is expressed as PS _n .

  Next, details of the storage unit 250 of the data processing unit 200 will be described with reference to FIG. Here, the configuration of the storage unit 250 necessary for processing related to the face detection function of the digital camera 1 will be described. That is, in order to realize the face detection function with the digital camera 1, a storage area as shown in FIG. 2B is secured in the storage unit 250 according to the present embodiment.

  As illustrated, the storage unit 250 includes storage areas such as a reference image storage area 250a, a reference image storage area 250b, and a program storage area 250c, and stores image data corresponding to each storage area. .

  The reference image storage area 250a and the reference image storage area 250b are storage areas for storing reference images that are referred to when face detection is performed. Here, the face detection function according to the present embodiment uses the face image prepared in advance as a reference image, as in the conventional face detection technology, and the feature portion indicated by the sub window image and the reference image set on the detection image. Is searched for a portion corresponding to the face on the detection image. The reference image storage area 250a and the reference image storage area 250b are prestored face images to be referred to at this time.

  In the present embodiment, two types of reference images indicating face images to be referred to in face detection are prepared and stored in the reference image storage area 250a and the reference image storage area 250b, respectively. An example of the reference image stored in the reference image storage area 250a is shown in FIG.

  As shown in the drawing, in the reference image storage area 250a, for example, a plurality of face images having a sub window size (for example, 24 × 24 dots) set at the time of detection are stored. Here, it is assumed that the reference image stored in the reference image storage area 250a is an image in which the face portion (feature portion) occupies almost the entire subwindow as shown in FIG. Such an image is a reference image (hereinafter referred to as “normal reference image FL”) that is normally used in the conventional face detection technology.

  In the reference image storage area 250b, a different type of reference image from the normal reference image FL is stored. In the present embodiment, an image that is the same face image as the normal reference image FL, but whose face part (feature part) is smaller than the normal reference image FL (hereinafter referred to as “reduced reference image FS”) is for reference. It is assumed that it is stored in the image storage area 250b. An example of such a reduced reference image FS is shown in FIG.

  As shown in the figure, the reduced reference image FS is an image having the same size as the sub-window size (for example, 24 × 24 dots), like the normal reference image FL. It is the image shown by the center part of the area | region. A plurality of such reduced reference images FS are stored in the reference image storage area 250b.

  The program storage area 250c is a storage area for storing an operation program executed by the control unit 210.

  In the present embodiment, the control unit 210 executes the operation program stored in the program storage area 250c of the storage unit 250, thereby realizing each process described later. In this case, the control unit 210 implements each process. The function will be described with reference to FIG.

  FIG. 5 is a functional block diagram showing functions realized by the control unit 210 when executing the functions according to the present invention. Note that the functions shown in FIG. 5 indicate functions that are required when the face detection function is executed.

  That is, when performing face detection on the finder screen (live view screen) at the time of imaging, the function shown in FIG. As illustrated, the control unit 210 includes an image conversion unit 211, a target image selection unit 212, an identification area setting unit 213, a first classifier 214, a second classifier 215, an identification result determination unit 216, and an output processing unit. 217, etc.

  The image conversion unit 211 converts the Rexro image output by the image sensor unit 120 for live view into an image used in face detection processing. That is, the VGA image stored in the input image storage area 230a is converted into a luminance image, stored as a detection through image PL in the detection image storage area 230b, and the detection through image PL is reduced for detection less than the QVGA size. The image is converted into an image PS and stored in the detection image storage area 230c. Note that the image conversion by the image conversion unit 211 is performed by a known image conversion algorithm or the like.

  The target image selection unit 212 selects an image to be processed for face detection. That is, one of the detection images stored in the detection image storage area 230b and the detection image storage area 230c is selected as a processing target image.

  The identification area setting unit 213 sets an identification area (subwindow) for identifying a face portion on the image selected by the target image selection unit 212. In this embodiment, it is assumed that a 24 × 24 dot sub-window is set on the selected screen. Further, when a face portion is detected in the subwindow set in the detection reduced image PS, the identification area setting unit 213 sets an extension area for setting the subwindow on the detection through image PL.

  The first discriminator 214 compares the subwindow image set by the discrimination area setting unit 213 with the reduced reference image FS (small size image) stored in the reference image storage area 250b, and the image of the subwindow. Identifies whether or not indicates a face portion.

  The second classifier 215 compares the image of the subwindow set by the identification area setting unit 213 with the normal reference image FL (normal size image) stored in the reference image storage area 250a, and the image of the subwindow. Identifies whether or not indicates a face portion.

  The first classifier 214 and the second classifier 215 according to the present embodiment are all known face detection classifier algorithms (for example, neural network (NN), Adaboost, support vector machine (SVM)). , Etc.) to identify the face portion. When performing face detection using such an algorithm, if an image other than the face portion is used as the reference image, it is stored in the reference image storage area 250a or the reference image storage area 250b as necessary. Shall.

  The identification result determination unit 216 determines the position of the face portion shown on the input Rexroh image by integrating the identification results obtained by the first classifier 214 and the second classifier 215.

  Based on the position of the face portion determined by the identification result determination unit 216, the output processing unit 217 displays the detected face portion clearly on the Rexroux image (live view image) displayed on the display unit 310. . Here, for example, a rectangular frame image as shown in FIG. 6 is displayed at the position of the detected face portion.

  The above is the function realized by the control unit 210 during the operation of the face detection function. In the present embodiment, each function described above is realized by a logical process performed by the control unit 210 executing a program. These functions are, for example, ASIC (Application Specific Integrated Circuit). Or an integrated circuit). In this case, among the functions shown in FIG. 5, the functions related to image processing may be realized by the image processing unit 220.

  The configuration of the digital camera 1 described above is a configuration necessary for realizing the present invention, and a configuration used for a basic function and various additional functions as a digital camera is provided as necessary. . In particular, regarding the configuration related to the face detection function, in this embodiment, what is necessary until a frame image is displayed on the detected face portion is shown. In the face detection function of the digital camera 1, detection is performed in this way. It is also assumed that an imaging operation in accordance with the face position is included. That is, it is assumed that a configuration and processing necessary for an operation of capturing an image by adjusting the focus and exposure to the detected face part are appropriately realized using a known face detection technique, and description thereof is omitted in the present embodiment.

  The operation of the digital camera 1 having such a configuration will be described below. Here, “face detection processing” executed by the digital camera 1 when imaging using the face detection function will be described with reference to the flowchart shown in FIG. This face detection process is started when the user of the digital camera 1 activates the face detection function of the digital camera 1 by operating the operation unit 330, for example.

  When the process is started, the image conversion unit 211 captures the input image (VGA Rexro image) stored in the input image storage area 230a (step S101), and converts it into a luminance image of the same size (that is, VGA). Then, it is stored in the detection image storage area 230b as a detection through image PL (step S102).

  The image conversion unit 211 converts the size of the detection through image PL stored in the detection image storage area 230b in step S101 from VGA to QVGA, and stores it as a detection reduced image PS in the detection image storage area 230c (step S101). S103).

  When the detection image having the same size (for example, VGA) as the input image and the detection image having the reduced size (for example, QVGA) are generated in this way, the image conversion unit 211 notifies the target image. Notify the selection unit 212. The target image selection unit 212 sequentially executes “enlargement direction face detection process (1)” and “enlargement direction face detection process (2)” in response to the notification from the image conversion unit 211 (steps S200 and S300). .

  Here, the enlargement direction face detection means that a face is detected by face identification using the reduced reference image FS on the reduced detection image (detection reduced image PS), and the detection through corresponding to the detected position is performed. Face recognition using the normal reference image FL is performed at a position on the image PL. That is, after face detection using a reduced size image is performed in the first stage of processing and the face position is estimated, face detection is performed in the subsequent stage based on the result on a normal size image whose image size is the enlargement direction. . In this case, the process in the former stage is “enlargement direction face detection process (1)”, and the process in the latter stage is “enlargement direction face detection process (2)”.

  The “enlargement direction face detection process (1)” executed here will be described with reference to the flowchart shown in FIG.

  As described above, since the reduced detection image is used in the preceding stage of the enlargement direction face detection process, the target image selection unit 212 stores the detection reduction reduced to QVGA stored in the detection image storage area 230c. The image PS is selected as a target image and read out to a work area developed in the RAM of the control unit 210 (step S201).

  When the target image selection unit 212 reads the target image, the target image selection unit 212 notifies the identification region setting unit 213 to that effect. In response to the notification from the target image selection unit 212, the identification region setting unit 213 sets an identification region (subwindow) in the reduced detection image PS read to the work area. In the present embodiment, as in the conventional face detection technique, an identification area having a predetermined size is set on the target image, and the face portion is detected while sequentially moving the identification area. Therefore, in such movement of the identification area, a movement route is defined in advance. In the present embodiment, for example, as shown in FIG. 11A, the upper left corner of the image is set as the start position, and the identification area SW is moved to the right by a predetermined step from there. When the identification area SW reaches the right end of the image, it is moved downward by one step and then moved leftward by one step. Then, every time the identification area SW reaches the edge of the image, the identification area SW is moved over the whole area of the image by repeating the operation of moving downward by one step and then moving in the reverse direction.

  Therefore, the identification area setting unit 213 according to the present embodiment sets, for example, an identification area SW having a size of 24 × 24 dots at the above-described identification start position (that is, the upper left corner of the image) (step S202). The detection reduced image PS is cut out in the set identification area SW (step S203).

  The identification area setting unit 213 inputs the clipped image to the second classifier 215 when the image of the set identification area SW is cut out. As described above, the second classifier 215 uses, for example, a face identification algorithm such as a neural network to compare the reduced reference image FS stored in the reference image storage area 250b with the input image. Identify whether the input image shows a face. Accordingly, the second discriminator 215 performs face identification on the input image cut out from the detection reduced image PS using the reduced reference image FS as shown in FIG. 4B (step S204).

  Here, for example, when a neural network is employed as the face identification algorithm, the second discriminator 215 outputs a teacher signal that is 1 for a face and −1 for a non-face. Such a teacher signal is based on a known neural network technique, and is set as needed by weighting by learning of the second discriminator 215 (first discriminator 214).

  When such a neural network is used as a face detection algorithm, a value close to 1 is output if the input image shows a face in face identification in step S204, and a value close to -1 is output if it is a non-face. Will be. The second discriminator 215 generates an output signal corresponding to the employed face detection algorithm and outputs the output signal to the discrimination result determination unit 216.

  Based on the output from the second discriminator 215, the discrimination result determination unit 216 discriminates whether or not a face is detected in the image cut out from the detection reduced image PS in the discrimination area SW (step S205).

  When a face is detected (step S205: Yes), the identification result determination unit 216 records the position of the identification area SW on, for example, a work area (step S206). Here, for example, a detection result table as shown in FIG. 11B is created on the work area, and the position of the identification region SW where the face is detected based on the reduced reference image FS on the detection reduced image PS. Is recorded. In this case, for example, coordinates on the detection reduced image PS indicating the center position of the identification area SW are recorded as position information of the identification area SW.

  On the other hand, when the identification result of the second classifier 215 indicates that the identification area SW is not a face (step S205: No), the identification result determination unit 216 does not record the position information of the identification area SW.

  In this way, when the identification result determination unit 216 determines the identification result, it notifies the identification region setting unit 213 to that effect. In response to the notification from the identification result determination unit 216, the identification region setting unit 213 determines whether or not the currently set position of the identification region SW is the identification end position on the detection reduced image PS (step S207). ) Until the identification at the end position is performed (step S207: No), the identification area SW is moved step by step (step S208). Here, the movement step of the identification area SW in the face identification in the detection reduced image PS is assumed to be, for example, 4 dots. That is, in the enlargement direction face detection process (1), while moving the identification area SW by 4 dots along the route as shown in FIG. 11A, a portion corresponding to each identification area SW is cut out, and the reduced reference image FS is extracted. Is used to identify the face (steps S203 to S206).

  Then, when the face detection operation at the end position is performed (step S207: Yes), this process ends, and the process returns to the face detection process flow shown in FIG. Here, since the first stage of the enlargement direction face detection process has been completed, the subsequent "enlargement direction face detection process (2)" is continuously executed. The enlargement direction face detection process (2) will be described with reference to the flowchart shown in FIG.

  When the process is started, the identification result determination unit 216 determines whether or not a face has been detected in the preceding enlargement direction face detection process (1) (step S200) (step S301). As described above, in the enlargement direction face detection process (1), when the face is detected, the position information of the identification area SW is recorded in the work area, so whether the position information of the identification area SW is recorded. By checking whether or not, it is possible to determine whether or not a face has been detected in the previous processing.

  Here, in the enlargement direction face detection processing according to the present embodiment, face detection using the reduced reference image FS with a small face portion is first performed on the detection reduced image PS that is smaller in size than the original live view image. Then, a portion that is estimated to be a face on the input image is given. For the detection through image PL having the same size as that of the actual live view image, face identification is not performed on the entire image area, but face identification is performed around the hit position.

  In the present embodiment, such a processing procedure is used, and thus the above-described enlargement direction face detection process (1) is a process up to hitting a portion estimated to be a face on the input image. The enlargement direction face detection process (2) that is currently being performed is a detailed face identification operation at the position hit in the previous process. Therefore, in the enlargement direction face detection process (1), when the part estimated to be a face is not detected, it is not necessary to perform the enlargement direction face detection process (2). Therefore, when the face is not detected in the preceding enlargement direction face detection process (1) (step S200) (step S301: No), this process is terminated as it is. In this case, the enlargement direction face detection process (2) (step S300) is not substantially executed.

  On the other hand, when a face is detected in the enlargement direction face detection process (1) (step S301: Yes), the identification result determination unit 216 notifies the target image selection unit 212 to that effect. In response to the notification from the identification result determination unit 216, the target image selection unit 212 selects the VGA size detection through image PL as a processing target, and reads it from the detection image storage area 230b to the work area (step S302).

  When the detection through image PL is read, the target image selection unit 212 notifies the identification region setting unit 213 to that effect. In response to the notification from the target image selection unit 212, the identification region setting unit 213 sets a narrowed detection region that is a region where a face detection operation is performed on the VGA detection through image PL (step S303). Here, by referring to the detection result table on the work area shown in FIG. 11B, the identification area setting unit 213 identifies the identification area SW in which the face has been detected in the preceding-stage enlargement direction face detection process (1). Is recognized (hereinafter referred to as “face detection portion”).

  Here, in the enlargement direction face detection process (1) in the previous stage, the detection reduced image PS reduced to QVGA is the processing target, so the identification region setting unit 213 performs relative detection of the face detection portion in the detection reduced image PS. A narrow detection area is set on the through image for detection PL read in step S302 based on the obtained relative position. That is, QVGA, which is the image size of the detection reduced image PS, is obtained by halving the length of each side of the VGA, which is the actual size of the live view image (area ratio 1/4). The reduced image PS for detection and the through image PL for detection are similar in shape. Therefore, if the relative positional relationship between one of the detection reduced images PS and the face detection portion detected there is obtained, the position corresponding to the face detection portion is also a relative position on the similar detection through image PL. Can be specified by relationship.

  The positional relationship between the reduced detection image PS and the detection through image PL will be described with reference to FIG. Here, a case where a captured image as shown in FIG. 12A is obtained will be described as an example. In the present embodiment, since the image size of the captured image is VGA, the image illustrated in FIG. 12A is a VGA detection through image PL. It is assumed that a human subject is imaged in such a VGA size image as shown in FIG. Here, as shown in the figure, it is assumed that the person who is the subject is relatively small compared to the size of the image.

  FIG. 12B shows an example in which such a captured image (detection through image PL) is reduced to a QVGA detection reduced image PS. Since the entire image is reduced, the size of the subject shown there is also relatively smaller than the detection through image PL shown in FIG.

  FIG. 12C shows an example of the identification region SW set when the above-described enlargement direction face detection process (1) is performed on such a detection reduced image PS. As described above, in the face identification in the enlargement direction face detection process (1), the face identification is performed using the reduced reference image FS as shown in FIG. In the present embodiment, since the size of the reduced reference image FS is the same size (24 × 24 dots) as the identification area SW, the identification area SW and the size of the subject are as shown in FIG. For example, as shown in FIG. 12D, a face is detected in the identification area SW at a plurality of positions according to the movement path of the identification area SW.

  The narrowed-down detection area set on the detection through image PL in step S303 is an area in which such a plurality of identification areas SW are collected. Therefore, when face detection is performed in a plurality of identification areas SW in the preceding enlargement direction face detection process (1), the narrowing detection area set in step S303 is at least the size of the identification area SW (this example) In this case, the area is larger than 24 × 24 dots. FIG. 13A shows an example in which such a narrowing detection region (hereinafter referred to as “narrowing detection region SS”) is set on the detection through image PL.

  That is, the identification area SW is set with the detection reduced image PS as shown in FIG. 12B, and the identification area SW when the face detection is performed at a plurality of positions as shown in FIG. If the detection area SS is set to the detection through image PL shown in FIG. 12A, the relationship between the size and the position of the subject image appearing in the detection through image PL and the narrowing detection area SS is shown in FIG. As shown in (a).

  FIG. 13B shows an enlarged view of the vicinity of the narrowing detection area SS set in this way. As shown in the drawing, when a plurality of identification areas SW in which faces are detected in the preceding enlargement direction face detection process (1) are integrated and set on the detection through image PL, narrowing detection is performed around the face portion of the subject. The area SS is set. Here, returning to FIG. 9, the explanation of the enlargement direction face detection process (2) will be continued.

  When the narrowing detection area SS as shown in FIG. 13A is set in step S303, the identification area setting unit 213 sets the identification area SW in the narrowing detection area SS (step S304). Here, an identification area SW having the same size (that is, 24 × 24 dots) as the identification area SW in the enlargement direction face detection process (1) described above is set. Here again, the identification area SW is moved along a route similar to the moving route shown in FIG. 11A, and the moving step is assumed to be 2 dots, for example.

  Here, since the movement path is indicated by an arrow in FIG. 11A, in step S304, a 24 × 24 dot identification area SW is set with the upper left corner of the narrowing detection area SS as a start position. . The identification area setting unit 213 cuts out a part corresponding to the identification area SW set in this way from the detection through image PL (step S305), and inputs it to the first classifier 214.

  As described above, the first discriminator 214 is a discriminator that performs face discrimination with reference to the normal reference image FL as shown in FIG. As shown in FIG. 4 (a), the normal reference image FL has a face portion that occupies the entire image having the same size as the identification area SW. Therefore, the VGA detection through image PL is converted from the QVGA detection reduced image PS. In this processing for performing face identification by enlarging the face, it is possible to perform face identification within the narrowed down detection area SS using such a normal reference image FL.

  Similarly to the second discriminator 215, the first discriminator 214 discriminates whether the clipped image is a face portion by a predetermined face discrimination algorithm, and outputs a signal indicating the discrimination result to the discrimination result determination unit 216. To do.

  The identification result determination unit 216 determines whether a face is detected based on the output signal from the first classifier 214 (step S307). If detected (step S307: Yes), the position information of the identification area SW on the detection through image PL is recorded in the work area (step S308), and if not detected, it is not recorded.

  Here, when the position information of the identification area SW is recorded in the work area, a detection result table as shown in FIG. 11C is created in the work area, and the position information of the identification area SW is recorded.

  When the identification result is determined in this way, the identification result determination unit 216 notifies the identification region setting unit 213 to that effect. In response to the notification from the identification result determination unit 216, the identification region setting unit 213 determines whether or not the current identification region SW is an identification end position in the narrowing detection region SS (step S309). Then, until the identification area SW that has performed face identification reaches an identification end position in the narrowing detection area SS (step S309: No), the identification area SW is sequentially moved by 2 dots (step S310), and the above-described The operations in steps S305 to S308 are repeated.

  That is, as shown in FIG. 13B, face identification is performed at each position while moving the identification area SW, and position information of the identification area SW when a face is detected is recorded. Here, as shown in FIG. 13B, on the VGA detection through image PL, the identification area SW of 24 × 24 dots is substantially the same as the size of the face portion of the subject, and the face identification is performed. In the normal reference image FL referred to in this case, since the face portion occupies the entire 24 × 24 dots, the face is detected in almost one identification area SW.

  In this way, when the face identification operation is performed over the entire narrowed detection area SS set on the VGA size detection through image PL (step S309: Yes), the identification area setting unit 213 informs the target image. The selection unit 212 is notified, and this process ends. In this case, the process returns to the face detection process flow shown in FIG.

  In the face detection process, a “reduction direction face detection process” (step S400) is executed following the enlargement direction face detection process (2). This reduction direction face detection process is a process generally performed in the conventional face detection technique, and generates an image pyramid (see FIG. 3) in which input images are sequentially reduced, and uses a reference image of a certain size. Thus, by performing face identification on each reduced image, the face can be identified even on an input image in which the size of the face portion within the angle of view is indefinite.

  In such a conventional face detection process, since the normal reference image FL as shown in FIG. 4A is used and face identification is performed with an image pyramid reduced to a size equal to or smaller than QVGA, the face portion with respect to the angle of view. The face portion can be detected when the size of is relatively large. Such reduction direction face detection processing will be described with reference to the flowchart shown in FIG.

  When the process is started, the target image selection unit 212 determines whether or not the enlargement direction face detection process (2) (step S300) has been executed (step S401). Here, when this process is executed after execution of the enlargement direction face detection process (2) (step S401: Yes), the detection through image PL (VGA image) is processed in the enlargement direction face detection process (2). Since the image is used as an image, the target image selection unit 212 selects the reduced QVGA detection reduced image PS as the processing target image in this processing, and reads it from the detection image storage area 230c (step S402). .

  On the other hand, when the face detection is not performed in the enlargement direction face detection process (1), as described above, the enlargement direction face detection process (2) is not substantially executed. In such a case (step S401: No), this process is executed after the enlargement direction face detection process (1) is completed. In the enlargement direction face detection process (1), since the detection reduced image PS (QVGA image) is used as the processing target image, the detection reduced image PS to be processed in this process has already been read. It will be.

The QVGA detection reduced image PS read first in this process is set as a detection reduced image PS ₀ (see FIG. 3). When the reading of the detection reduced image PS ₀ to be processed is completed, the target image selection unit 212 notifies the identification region setting unit 213 to that effect. The identification area setting unit 213 sets the identification area SW with the upper left corner of the detection reduced image PS ₀ as the start position by the same process as the enlargement direction face detection process (1) described above (step S403). cutting out portions corresponding to the SW from the detection reduced image PS ₀ (step S404).

Here, unlike the above-described enlargement direction face detection process (1), the identification area setting unit 213 inputs the cut-out image to the first classifier 214. Thus, the face identification in the image extracted from the detection reduced image PS ₀ of the reduced QVGA, usually performed using the reference image FL as shown in FIG. 4 (a) (step S405).

  Based on the output signal from the first discriminator 214 that has performed face discrimination in this way, the discrimination result determination unit 216 determines the discrimination result (step S406). If a face is detected (step S406: Yes), the current position information of the identification area SW is recorded in the work area (step S407). When such a determination operation is performed, the identification result determination unit 216 notifies the identification region setting unit 213 to that effect.

The identification area setting unit 213 moves the identification area SW by two dots every time the identification result is determined (step S408: No, step S409). Then, by performing the processing of step S404~ step S405 at the position of each identification area SW, performs face identification across the detection reduced image PS _0, the position information of the identification area SW a face has been detected in the work area Record. Here, a detection result table as shown in FIG. 11D is created in the work area, and the position information of the identification area SW where the face is detected is recorded.

When the face identification operation in the entire detection reduced image PS ₀ is completed (step S408: Yes), the identification result determination unit 216 notifies the image conversion unit 211 to that effect. Image converting unit 211 in response to the notification from the identification result determining unit 216, and stores the detected image storage area 230c by reducing the detection reduced image PS ₀ being currently processed (step S410). Here, for example, to reduce the size of the detected thumbnail images PS ₀ to 80%.

  When the image conversion unit 211 reduces the image in this way, the image conversion unit 211 notifies the target image selection unit 212 to that effect. In response to the notification from the image conversion unit 211, the target image selection unit 212 determines whether the size of the reduced detection reduced image PS is larger than the size of the identification region SW used for face identification (that is, 24 × 24 dots). Is determined (step S411).

  If the size of the reduced detection image PS after reduction is larger than the size of the identification area SW (step S411: No), the above-described processing from step S403 to step S409 is performed on the reduced detection reduction image PS. That is, face recognition is performed using the normal reference image FL (FIG. 4A) while moving the identification area SW on the reduced detection reduced image PS.

Then, each time the face identification operation is performed in the entire area of the reduced detection reduced image PS, the detection reduced image PS is further reduced (step S408: Yes, step S410). In other words, image pyramids as shown in FIG. 3 are sequentially generated until the size of the reduced detection reduced image PS reaches the size of the identification area SW, and in each detection reduced image PS _x (x = 0 to n), A face identification operation using the normal reference image FL is performed.

  In such processing, face recognition is performed while moving the identification area SW on the detection reduced image PS. Therefore, if the detection reduced image PS itself is smaller than the identification area SW, the identification operation cannot be performed. Therefore, when the reduced detection image PS after reduction becomes smaller than the identification area SW (step S411: Yes), this process ends, and the process returns to the face detection process flow shown in FIG.

  In the face detection process (FIG. 7), the identification area SW when the face is detected in each of the enlargement direction face detection process (1), the enlargement direction face detection process (2), and the reduction direction face detection process described above. Since the position information is recorded in the work area, the identification result determination unit 216 integrates the detection results in the enlargement direction face detection process (2) and the reduction direction face detection process (step S104).

  That is, in the captured image, the detection result of the enlargement direction face detection process (2) for detecting a face portion that is relatively small and the reduction direction face detection process for detecting a face portion that is relatively large Can be identified even if a plurality of faces appear in various sizes in the captured image.

  Here, when both the enlargement direction face detection process (2) and the reduction direction face detection process are performed, the face identification operation is performed with a plurality of image sizes for the same captured image. Is represented by a frame image to be finally displayed, as shown in FIG. 14, a plurality of frames are concentrated on the face portion. Here, the identification result determination unit 216 determines that a portion where a certain number or more of detection results are obtained at the same position is a face portion, considering that there is a misrecognition at a certain ratio, and at that position. By integrating these detection results into one, a live view screen as shown in FIG. 6 can be obtained.

  In this case, the identification result determination unit 216 notifies the output processing unit 217 of position information based on the integrated result. The output processing unit 217 controls the image output unit 240 in response to the notification from the identification result determination unit 216, so that a live view image with a face detected as shown in FIG. Step S105).

  Then, by repeating the processing from step S101 to step S105 until the end event of the face detection operation occurs (step S106: No), the frame image is always displayed on the face portion when the face is reflected on the live view image. Will be displayed.

  Here, for example, when the end event of the face detection operation such as pressing of the shutter button or the end of the shooting mode is generated (step S106: Yes), this processing is ended. When the shutter button is pressed, processing for performing imaging operation and image processing (for example, focusing operation, exposure adjustment, white balance adjustment, etc.) according to the detected face portion is separately executed. .

  As described above, by applying the present invention as in the above-described embodiment, even when a person is photographed as a subject, the face portion appearing in the captured image is surely small. Face detection can be performed.

  In this case, in addition to the reference image used in the conventional face detection technique, a reduced reference image having a face portion smaller than that of the normal reference image is prepared, and the image is reduced to QVGA or the like. By performing face identification using the reduced reference image, even when a relatively small face is shown in the captured image, it is possible to keep the position around that position. Then, only when the hit can be achieved, face identification using a normal reference image is performed on an image having the same size as the actual Rexroux image.

  Here, when the identification operation is performed with an image of an actual size, the operation range is narrowed down at the position where the hit is made, so that the face detection can be performed with a small processing load. In other words, as the size of the target image increases, the number of processing steps increases and the processing load increases when the identification operation is performed over the entire area. Even if the size of the image to be actually output is large, the identification operation is performed within the narrowed down range, so that the number of processing steps can be reduced as compared with the case where the entire area is identified. As a result, face detection can be performed without reducing the processing speed even when the image size increases.

  Here, in the conventional method of performing face identification with an image pyramid obtained by sequentially reducing the target image, the size of the face portion appearing on the first target image is the size of the face portion indicated in the reference image. However, the face detection cannot be performed unless it is larger than the above, but by performing the enlargement direction face detection processing exemplified in the above embodiment, even if the face shown in the captured image is relatively small, it is ensured. Can be detected.

  Furthermore, by performing such an enlargement direction face detection process and a reduction direction face detection process similar to the conventional process, for example, when shooting a plurality of subjects with different distances from the imaging device, various Even when a large-sized face appears on one captured image, face detection can be reliably performed.

  Therefore, the enlargement direction face detection process is increased in comparison with the conventional process, but a more reliable face detection can be realized by minimizing an increase in processing load.

  In the enlargement direction face detection process (1) in the above-described embodiment, a reduced image with a low processing load is used. If a hit cannot be made here, an enlargement direction face detection process (2) using a non-reduced image. ) Is not executed. That is, since only the necessary face detection operation is performed according to the content of the captured image, more reliable face detection can be realized while suppressing an increase in processing load to a minimum.

  The said embodiment is an example and the application range of this invention is not restricted to this. That is, various applications are possible, and all embodiments are included in the scope of the present invention.

  First, in the above-described embodiment, the case where a face is detected at the time of photographing with a digital still camera is illustrated, but the detection target is not limited to a face and is arbitrary. That is, the present invention can be applied to detection of various objects by preparing a reference image corresponding to a desired detection object.

  In this case, the detection target is not limited to the subject at the time of imaging. That is, since it is only necessary that the target image is included in the input image, the captured image may be used as the input image instead of performing detection at the time of imaging as exemplified in the above embodiment. That is, the configuration and functions of the data processing unit 200 shown in the above embodiment do not need to be installed in the imaging apparatus, and may be realized as various apparatuses that can input image data.

  Therefore, it is realizable as an image processing apparatus provided with the structure and function concerning this invention. Such an image processing apparatus can be realized by a general-purpose computer apparatus such as a personal computer as well as being realized as a dedicated apparatus having the configuration and functions according to the present invention in advance.

  In this case, a program similar to the program executed by the control unit 210 shown in the above embodiment is applied to a general-purpose computer apparatus, and the computer apparatus executes the program, thereby functioning as an image processing apparatus according to the present invention. Can be made.

  Further, when the present invention is realized by an imaging apparatus such as the digital camera 1 exemplified above, it can be provided as an imaging apparatus having the configuration and functions according to the present invention in advance, and by applying a program, The imaging device can also function as the imaging device according to the present invention.

  In the above embodiment, a digital still camera is shown as an example of an imaging device. However, the present invention is applied to an imaging device that performs moving image imaging, such as a digital video camera, as long as the captured image can be digitally processed. May be. In addition, the form of the image pickup apparatus is arbitrary, and it can be realized by a single digital still camera or digital video camera. In addition, the image pickup apparatus can be used in various electronic devices (for example, cellular phones) having the same image pickup function. The present invention may be applied.

  Even in such a case, by applying the program, an existing apparatus can be caused to function as the imaging apparatus according to the present invention.

  The application method of such a program is arbitrary. For example, the program can be applied by being stored in a storage medium such as a CD-ROM or a memory card, or can be applied via a communication medium such as the Internet.

  Moreover, in the said embodiment, although VGA and QVGA were shown as an example of image size, these are examples and image size is not restricted to these. That is, when performing the processing according to the present invention, the image size to be reduced and converted may be smaller than the input image size. Further, the size ratio between the input image and the reduced image is also arbitrary, and may be set as appropriate so that the processing load is optimized.

  Also, the algorithm employed when detecting the target image is arbitrary, and is limited to the neural network, Adaboost, support vector machine, etc. shown in the above embodiment as long as the desired target can be identified. It is not a thing.

It is a block diagram which shows the structure of the digital camera concerning embodiment of this invention. FIG. 2 is a diagram for explaining the image memory and the storage unit illustrated in FIG. 1, in which (a) illustrates an example of a storage area defined in the image memory, and (b) illustrates an example of a storage area defined in the storage unit. Show. It is a figure which shows the example of the image for a detection stored in the image storage area for a detection shown to Fig.2 (a). FIG. 3 is a diagram illustrating an example of a reference image stored in the reference image storage area illustrated in FIG. 2B, where FIG. 2A illustrates an example of a normal reference image, and FIG. 2B illustrates an example of a reduced reference image; . It is a functional block diagram which shows the function implement | achieved by the control part shown in FIG. It is a figure which shows the example of the live view image displayed on the display part shown in FIG. It is a flowchart for demonstrating the "face detection process" concerning embodiment of this invention. 8 is a flowchart for explaining an “enlargement direction face detection process (1)” executed in the face detection process shown in FIG. 7. It is a flowchart for demonstrating the "magnifying direction face detection process (2)" performed by the face detection process shown in FIG. It is a flowchart for demonstrating the "reduction direction face detection process" performed by the face detection process shown in FIG. It is a figure for demonstrating the operation | movement in each process shown in FIGS. 7-10, (a) shows the example of the movement path | route of an identification area, (b) is produced by the expansion direction face detection process (1). (C) shows an example of the detection result table created by the enlargement direction face detection process (2), and (d) shows the detection result table created by the reduction direction face detection process. An example is shown. FIG. 10 is a diagram for explaining the operation in the enlargement direction face detection process (2) shown in FIG. 9, where (a) shows an example of a detection through image, (b) shows an example of a detection reduced image, (c) shows an example of an identification area set on the detection reduced image shown in (b), and (d) shows an example of a face detection position by the identification area shown in (c). FIG. 10 is a diagram for explaining an operation in the enlargement direction face detection process (2) shown in FIG. 9, where (a) shows an example of a narrowing detection area set in the detection through image, and (b) shows (a The example of the identification area | region set in the narrowing-down detection area | region shown in FIG. It is a figure for demonstrating the concept of the operation | movement which integrates a detection result in the face detection process shown in FIG.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 ... Digital camera, 100 ... Imaging part, 110 ... Optical apparatus, 120 ... Image sensor part, 200 ... Data processing part, 210 ... Control part, 211 ... Image conversion part, 212 ... Target image selection part, 213 ... Identification area setting 214, first discriminator, 215, second discriminator, 216, discrimination result determination unit, 217, output processing unit, 220, image processing unit, 230, image memory, 230a, input image storage area, 230b. ... detection image storage area, 230c ... detection image storage area, 240 ... image output section, 250 ... storage section, 250a ... reference image storage area, 250b ... reference image storage area, 250c ... program storage area, 260 ... External storage unit, 300 ... interface unit, 310 ... display unit, 320 ... external interface unit, 330 ... operation unit, PL ... through image for detection, PS ... Out for a reduced image, FL ... usually the reference image, FS ... reduced reference image, SW ... identification area, SS ... narrowing the detection area

Claims (8)

In an image processing apparatus for detecting a detection target image on an input image by an identification operation using a reference image,
Reference image storage means for storing the reference image;
Reduced reference image storage means for storing a reduced reference image obtained by reducing the characteristic portion indicated by the reference image stored in the reference image storage means;
Image reduction means for reducing the input image;
Target position specifying means for specifying the position of the detection target image in the reduced image by performing an identification operation using the reduced reference image on the image reduced by the image reducing means;
Image detecting means for detecting the detection target image on the input image by performing an identification operation using the reference image at a position on the input image corresponding to the position specified by the target position specifying means. When,
An image processing apparatus comprising: The image detection unit does not perform an identification operation on the input image when the target position specifying unit cannot specify the position of the detection target image on the reduced image.
The image processing apparatus according to claim 1. The image reduction means sequentially reduces the input image,
The image detection means detects the detection target image by performing an identification operation using the reference image in each of the images sequentially reduced by the image reduction means.
The image processing apparatus according to claim 1, wherein the image processing apparatus is an image processing apparatus. In an imaging device that detects a detection target image on an input image by an identification operation using a reference image,
Input image acquisition means for acquiring a captured image as an input image;
Reference image storage means for storing the reference image;
Reduced reference image storage means for storing a reduced reference image obtained by reducing the characteristic portion indicated by the reference image stored in the reference image storage means;
Image reduction means for reducing the input image;
Target position specifying means for specifying the position of the detection target image in the reduced image by performing an identification operation using the reduced reference image on the image reduced by the image reducing means;
Image detecting means for detecting the detection target image on the input image by performing an identification operation using the reference image at a position on the input image corresponding to the position specified by the target position specifying means. When,
An imaging apparatus comprising: The image detection unit does not perform an identification operation on the input image when the target position specifying unit cannot specify the position of the detection target image on the reduced image.
The imaging apparatus according to claim 4. The image reduction means sequentially reduces the input image,
The image detection means detects the detection target image by performing an identification operation using the reference image in each of the images sequentially reduced by the image reduction means.
The imaging apparatus according to claim 4 or 5, wherein An image processing method for detecting a detection target image on an input image by an identification operation using a reference image,
An image reduction step for reducing the input image;
An object for specifying the position of the detection target image in the reduced image by performing an identification operation using the reduced reference image obtained by reducing the characteristic portion indicated by the reference image on the image reduced in the image reduction step. A location step;
An image detection step of detecting the detection target image on the input image by performing an identification operation using the reference image at a position on the input image corresponding to the position specified in the target position specifying step. When,
An image processing method comprising: A program for causing a computer to detect a detection target image on an input image by an identification operation using a reference image,
In the computer,
A function of reducing the input image;
A function for identifying the position of the detection target image in the reduced image by performing an identification operation using the reduced reference image obtained by reducing the characteristic portion indicated by the reference image on the reduced image;
A function for detecting the detection target image on the input image by performing an identification operation using the reference image at a position on the input image corresponding to the specified position;
A program characterized by realizing.

Images